A Multi-Radio Access Architecture for Ambient Networking

Transcription

1 A Multi-Radio Architecture for ing Johan Lundsjö 1), Ramón Agüero 2), Eftychia Alexandri 3), Fredrik Berggren 4), Catarina Cedervall 5), Konstantinos Dimou 6), Jens Gebert 7), Ralf Jennen 8), Ljupco Jorguseski 9), Reza Karimi 10), Francesco Meago 11), Haitao Tang 12), Riccardo Veronesi 13) 1) Ericsson (Sweden), 2) University of Cantabria (Spain), 3) France Telecom R&D (France), 4) Royal Institute of Technology (Sweden), 5) TeliaSonera (Sweden), 6) Panasonic European Laboratories (Germany), 7) Alcatel SEL (Germany), 8) RWTH Aachen University (Germany), 9) TNO Telecom(The Netherlands), 10) Lucent(UK), 11) Siemens (Italy), 12) Nokia (Finland), 13) Consorzio Ferrara Ricerche, University of Ferrara (Italy) Abstract The s concept targets forthcoming dynamic communication environments; characterized by presence of a multitude of different wireless devices, network operators and business actors that can form instant inter-network agreements with each other. In this paper we propose an architecture, including the concepts of Multi- Radio Resource Management and a Generic Link Layer, for efficient use of heterogeneous access technologies in such emerging scenarios. I. INTRODUCTION ODAY there are many different radio access Ttechnologies (RATs) that differ in their support of data rates, mobility, coverage, quality of service, and possible business models. In the future, additional RATs will be available with other characteristics supporting new challenging networking scenarios, but most likely not replacing the existing technologies. The mixture of heterogeneous RATs will therefore be an important characteristic of future wireless communications. There is plenty of prior research on how to combine different RATs, including a number of IST FP projects, e.g. BRAIN, MIND, DRiVE, WINE, ARROWS, MONASIDRE and EVEREST. However, this research has only tackled partial issues towards full network collaboration at the radio access level. The architecture presented in this paper is a result of the research on multi-radio access () within the s (AN) project [1]. It extends and generalises previous work to apply to any existing and future RATs. The architecture is novel in its provisioning of functionalities and mechanisms to support the AN vision of a dynamic environment with a multitude of different wireless devices, network operators and business actors that can form instant inter-network agreements with each other. These inter-network agreements can be the result of network collaboration or competition. The AN proposes a consistent framework and a complete architecture that on one hand allows joint radio resource management across different RATs and on the other hand enables new business relationships among involved actors. The paper is structured as follows: in Section II we give a brief overview of the s concept. Motivation, drivers and requirements for are discussed in Section III. Section IV gives an overview of a proposed architecture, followed by more specific details of its main components Multi-Radio Resource Management () and Generic Link Layer () in Sections V and VI respectively. Some critical architecture design options are summarized in Section VII. Finally Section VIII concludes the paper. More thorough descriptions of the overall AN concept, the, and the are provided in [2], [3] and[4]. II. THE AMBIENT NETWORKS CONCEPT An can be seen as a new type of network-level building block, existing above the level of individual devices or functions. It consists of a set of one or more nodes and/or devices, which share a common control plane, and implement well-defined external interfaces to Users or other ANs. The core concept of the AN architecture is illustrated at its highest level in Figure 1. s include a flexible set of control plane functions, which together comprise the Space (ACS), including the multi-radio access functions that are the main focus of this paper. The underlying data transfer and other user plane capabilities of the existing or new networks are accessed and controlled through an Resource (ARI). Together, they expose an Service (ASI) to upper layer services and applications running within the network. Inter-network cooperation between different ACSs is mediated through an integrated set of protocols at the (ANI). A basic mechanism of the AN concept is the dynamic and instant composition of networks without the need for preconfiguration or offline negotiation between network operators. Composition between ANs, which takes place over the ANI, enables the usage of resources without the need for long-term subscriptions, optimizes the delivery path and allows operators to flexibly integrate network technologies. Routing Group Information Context Information Agreement Establishment Traffic Engineering Service Space Connectivity Figure 1: Logical view of an Multi-Radio Overlay Support Layer Naming Resource

2 (1) Single-Operator (2) Cooperating -Operators (3) Multi-Hop and Local s AN2 Prov ider AN2 AN3 WLAN Figure 2: Three scenario examples AN4 Relaying AN2 WiMAX AN5 Relaying AN6 Local III. MULTI-RADIO ACCESS; MOTIVATION, CONCEPTS AND EXAMPLE SCENARIOS A. Motivation for the Architecture Drivers for new functionality can be identified from the perspective of end users, providers and regulators. Users and providers have a common interest when it comes to flexible use of different types of wireless accesses, including selection of a best type of access, both from a user point of view (e.g. low cost versus high performance) as well as from a provider s perspective (e.g. load sharing). This calls for the capability of overall management of network resources of multiple network operators supporting required service provisioning. Users will expect service continuity when moving between accesses, sometimes (depending on application) implying the capability of lossless and fast handover between different Radio es (RAs). The term RA is here used to refer to uncoupled radio channels, either across different RATs or within a single RAT (e.g. associated to different access providers). Furthermore, users will benefit from getting access to any network, requiring support for rapid establishment of roaming agreements (dynamic roaming) and efficient announcing strategies (of both user needs and provider offers). Given the multitude of already existing, emerging and future RATs it should be in the providers interest that there are means for efficient and relatively simple migration from present networks, to networks with more capabilities and support for more RATs. Another common area of interest for users and providers is reduced cost, which may be enabled by new flexible deployment concepts [5]. The interests of regulators include means for increased competition on the market and efficient use of spectrum. B. Multi-Radio Features and Concepts The architecture consists of two main components: - Multi Radio Resource Management (), for joint management of radio resources and load sharing between the different RAs. - Generic Link Layer (), which provides unified link layer processing, offering a unified interface towards higher layers and an adaptation to the underlying RATs. These are built on previous research, such as [7][8][9] and [10][11] respectively, generalised and extended with WL AN functionality for dynamic ing. A main feature of the architecture is resource sharing and dynamic agreements between ANs, including different access providers, through composition. Other novel features include efficient advertising, discovery and selection of RAs, including the possibility for a user to simultaneously communicate over multiple RAs, in parallel or sequentially. Furthermore, the architecture includes support for multi-radio multi-hop communication, including moving and fixed relays. C. Scenario Examples The benefits and novelty of the architecture can be exemplified by the three cases of AN configuration depicted in Figure 2. Each of these cases takes advantage of, but with some differences in the involvement of the functionality. The cases could also be seen as describing the time aspect and thus reflect migration aspects towards AN. In the first case a multi-radio access capable terminal () is connected to a single operator (AN2). Note that, in accordance with the AN philosophy, a terminal including an ACS is regarded as an AN of its own. This first case is to some extent already solved by state of the art solutions, e.g. national roaming between RAs belonging to one operator, and load management possibility for the specific combination of WCDMA-GSM [6]. The proposed architecture adds generic resource allocation functionality that enables load management over all possible RAs, and instant mapping of data flows to different RAs. The second case goes one step further by adding the possibility for information exchange between ANs belonging to different operators (AN2 and AN3), allowing terminals to access these separate ANs in a seamless manner. The level of information to be exchanged depends on the business relationship between the operators, ranging from full competitors to a cooperation relation. The third case adds multi-hop and local access provider concepts. An operator (AN2) allows a user terminal () to establish connectivity in three additional ways: via another terminal acting as relay (AN4), via a set of fixed relay nodes (AN5), or through a local access point (provided by local AN6 access provider). signalling and user data may be separated on different RAs. An example could be that a local access provider (AN6), at a certain time instant, provides a direct user data connection to the terminal, whereas a cooperating wide area coverage access provider (AN2) handles the associated control signalling towards the terminal. In all these cases the architecture offers means for efficient resource sharing, within and between ANs. D. Architecture Design Trade-Offs There are many issues that affect design choices and trade-offs have to be made. In the following a number of critical factors that have been considered in the architecture design are exemplified. The time scale of operation affects e.g. on what level in the protocol stack and how close to the radio interface certain functions are implemented. For example, the dynamic process of selecting the best RA may be based on

3 several metrics that vary over time. The time scale at which these metrics vary may differ a lot. At the extreme the RA selection may follow fast fading for one link (in the order of milliseconds), whereas if only system load is to be considered it may be enough to operate on a time-scale in the order of seconds. These aspects have been considered when proposing different levels of integration of RAs, and a functional split between and. Signalling overhead and Signalling delay have impact on the choice of which type of protocol a certain function shall employ, at which level of the protocol stack it should be situated, and whether the protocol termination needs to be centralized or distributed in the network. There is naturally a dependency to the time scale of operation. Deployment cost needs to be considered, e.g. when explicitly ensuring support for multi-hop networks and new deployment concepts. Migration aspects need to be taken into account, allowing gradual introduction of features and thereby evolve legacy networks into fully AN-compliant networks. The type and amount of information that can be exchanged between different ANs can be restricted for technical or non-technical reasons. For instance, a party (e.g. an operator) may not be willing to reveal certain information that may be relevant to a certain function that is operated by another party. IV. MULTI-RADIO ACCESS ARCHITECTURE OVERVIEW A high-level view of the proposed architecture is illustrated in Figure 3, showing functional blocks in a layered model, including user plane data flow (solid lines) and ( and ) signalling (dashed lines) through the layers. Arrows indicate control interfaces between different functional blocks, carrying information exchange and control commands e.g. for configuration or for measurement data retrieval. Note that only one communicating peer (network or terminal) is depicted. For the single-hop case the model can simply be mirrored at the other end. The is introduced on top of, and is partly replacing, the RA specific parts of the link layer. The toolbox of link layer functions within the provides a unified interface towards upper layers ( and above) in the user plane and provides adaptation towards the underlying (remaining RAT specific) link layers. The functions are built upon, or mapped onto the network intrinsic RRM functions, which belong to the underlying RA and are therefore not within the explicit scope of the AN. Two possibilities are illustrated for how signalling is conveyed between communicating entities, either over or directly mapped onto the. Finally, the figure illustrates information exchange between, and other ACS functions, here exemplified by mobility control and connectivity control. The model suggests a functional split between and the based on the design trade-offs described in Section III.D. In general the encompasses functions that are located close to the user plane of a data flow and/or need to operate on a relatively fine time scale. One example is the RA selection for which a hierarchical distribution of functionality between and is proposed, where the dynamically handles the mapping of data flows to any of the RAs selected by. Another example is that provides and reuses context information that is transferred between entities at RA reselection for seamless access switching. The type of context information depends on the scenario, e.g. depending on the types of RAs and if they are located within the same or different ANs. V. MULTI-RADIO RESOURCE MANAGEMENT A. Functionality Triggered by events such as session arrivals, mobility and resource optimisation, operates at system, session and flow level. At the system level, performs e.g. spectrum, load and congestion control. At the session level, coordinates decisions on different associated flows. At the flow level, establishes and maintains RAs that are possibly constituted of parallel multi-hop routes. Two main types of functions can be distinguished. The RA coordination functions are generic functions providing coordination abilities that span over the available RAs. The complementing RRM functions provide missing, or complement inadequate, RRM functions to an underlying RAT such as admission control, congestion control etc. They are also responsible for providing a generic interface to the RA coordination functions through adaptation towards the -intrinsic RRM functions. In the remainder of this section we concentrate on the RA coordination functions. The following definitions will facilitate the description: - Detected Set ( DS) is the set of all RAs that have been detected by through e.g. scanning or reception of RA advertisements. - Candidate Set ( CS) is the set of RAs that are candidates to be assigned by to a given data flow; it is always a flow-specific subset of the DS. - Active Set ( AS) is the set of RAs assigned by to a given data flow at a given time, and is always a subset of the CS. - Active Set ( AS) is the set of RAs assigned to a given entity by to serve a given data flow at a given time; it is always a subset of the AS Information included in each set may cover RA identity, capabilities, related measurements, access costs etc. The set definitions are valid in single-hop as well as multi-hop cases, combining multiple RAs sequentially (multi-hop) and/or in parallel. For scalability reasons, and as a possible result of composition, the different sets are defined recursively: e.g. an entity may expose an AS as a single RA, hiding its inner complexity. RA Advertising informs about the presence of a network or its capabilities to provide a given service possibly in a business oriented fashion (with associated costs). For example, proxy advertisements could be sent on behalf of other access providers or network nodes.

4 Other ACS Functions Mobility contr ol Connectivity Intrinsic RRM Figure 3: High-level functional layer architecture RA Specific Link Layer The RA Discovery function may use the RA Advertisements to identify and monitor candidate RAs and routes for specific flows. Thereby it establishes and maintains the DS and CS. The RA Selection function selects the appropriate RAs for a given flow, thereby establishing the AS and the respective AS. The first step of a RA Selection process is the RA Evaluation wherein several parameters may be considered, including signal quality and strength, end-user QoS needs, end-user cost-capacity performance, multioperator network capacity, RA capabilities, RA status, RA availability, user and provider preferences and policies, and operator revenues in single/multi-operator scenarios. The evaluation is then followed by an RA Admission decision, ensuring that already established QoS agreements are protected. Negotiation includes negotiation of roles during composition, and exchange of relevant information during operation, through various forms of information exchange. At the system level, Overall Resource Management keeps a global control of network resources and protects established QoS agreements proactively within an AN and in coordination with other ANs. Means for this include load sharing, excess QoS elimination, QoS downgrading, flow/session dropping and dynamic spectrum control within or between RAs. B. Distribution functionality may be distributed within and across ANs in various ways. This will affect the type and amount of information that can be exchanged, and thus the possible degree of coordination, which in turn relates to the business models implied by the AN cases. Flexibility is provided by the capability of entities to negotiate their respective roles towards each other. Centralized control is likely to be beneficial in any network composed by a number of ANs belonging to a same administrative entity e.g. a personal area network (PAN). Another example is a centralized entity acting as an access broker, delivering commands or policies between ANs, or simply handling exchange of limited information such as availability of RAs. Distributed solutions could be used in large networks for scalability reasons, or when a central coordination is simply not desirable due to the fact that the involved administrative entities pursue different strategies or don t trust each other. Scenarios characterized by a large number of small networks without a-priori relations are believed to put high requirements on the negotiation of roles. Figure 4: Four levels of integration RLC -RLC -RLC RLC U- plane -RLC - VI. GENERIC LINK LAYER Conceptually, the consists of a toolbox of functionalities, which provides a unified interface to higher layers and facilitates efficient link-layer inter-working among multiple, possibly diverse, RAs. Along with the generic toolbox, the proposed enables two novel applications. The first of these, named Multi-Radio Transmission Diversity (MRTD), implies the sequential or parallel use of multiple RAs for the transmission of a traffic flow. The second, termed Multi-Radio Multi-Hop ing (MRMH), implies potential link layer support for multiple RAs along each wireless connection over a multi-hop communication route. A. Toolbox The has been defined as a toolbox of functions, making it possible to construct several alternative or complementing solutions depending on design tradeoffs. Figure 4 shows four different levels of integration that have been defined. A major factor determining these levels has been the possible time scale of operation. The further down the extends in the link layer, the finer the time scale of operation can be. For instance, the lower right case where is integrated all the way down to the level allows for dynamic scheduling onto different RAs on a per- PDU basis. These different levels offer a migration strategy. The model separates the into the functional blocks -C (control), -D (data), -RLC and -. -C contains functions for the link layer configuration and the interaction with, including Selection, Resource Monitoring and Performance Monitoring functions. Furthermore, it provides information on the QoS and the demands of different traffic flows. -C negotiates with the AS, within which may autonomously perform RA selection. In addition, supports mobility management functions and facilitates context transfer (-D) and security management (-

5 D, -RLC). The functionality for support of MRTD and MRMH include Scheduling (-D, -), Buffer Management (-D), Error and Flow, Segmentation and Reassembly (-RLC). B. Multi-Radio Transmission Diversity Based on a novel extension of the various transmission diversity mechanisms applied in legacy systems, MRTD may be broadly defined as the dynamic selection of multiple RAs for the transmission of a user s data. MRTD can be performed at different protocol levels ( or level), with support of different RA re-selection rates. In addition, multiple RAs can be used for parallel (simultaneous) or switched (sequential) transmission of a traffic flow. Different MRTD schemes can also be combined. Typically, different RAs integrated in a common node can apply level MRTD, which enables high RA re-selection rates by exploiting instantaneous feedback from radio link quality. However, this would require sharing detailed control information between the two ANs, which may not be possible or desirable for all business models implied by e.g. scenarios 2) and 3) described in Section III.C. In these situations level integration would probably be the preferred alternative. C. Multi-Radio Multi-Hop A potential feature of the is Multi-Hop ARQ spanning over the complete multi-hop route, which may be described in terms of a two-stage error recovery process. The principle is that an intermediate relay node responds with an acknowledgement to the previous node when a data packet has been successfully received, thereby turning into a passive ARQ state. However, the previous node (and all nodes along the route) will not delete the transmitted data from its ARQ buffer until reception of a final acknowledgement from the final receiver. Associated functions include Adaptation to different RATs on different hops to handle different segmentation sizes per hop, and flow control and priority based queuing along the route. MRMH can be combined with MRTD. Henceforth two route selection mechanisms can be identified, those that address the problem within the route (i.e. at the relay nodes) and those that address it from the edge-nodes of the network (i.e. infrastructure nodes or user terminals). VII. ARCHITECTURE DESIGN OPTIONS The flexibility of the proposed architecture implies that many design choices have to be made for the implementation. Critical options, where the trade-offs discussed in section III.D have fundamental implications, include: - Centralized versus distributed coordination between multiple ANs, as discussed in Section V.B. - signalling carried by or directly mapped onto, as indicated in Section IV. - MRTD on packet level versus PDU level, as discussed in Section VI.B - Multi-radio multi-hop with or without explicit support from and/or as discussed in Sections V and VI.C respectively. Note that choices do not necessarily need to be strict, i.e. combinations of alternatives may very well exist in the same implementation. Feasibility and performance evaluations of various options are currently being evaluated. VIII. CONCLUSION We have presented a Multi-Radio architecture, supporting the vision of a dynamic environment with a multitude of different wireless devices, network operators and business actors that can form instant inter-network agreements with each other. The main components of the architecture are Multi-Radio Resource Management, providing joint management of radio resources, and the Generic Link Layer, which offers a unified interface towards higher layers and an adaptation to the underlying accesses. We have discussed the various features proposed, along with scenarios, requirements and design options. Feasibility and performance of the proposed architecture and its features are currently under study and will be presented in subsequent publications. ACKNOWLEDGMENT This paper describes work undertaken in the context of the s project which is part of the EU s IST program. In total 41 organizations from Europe, Canada, Australia and Japan are involved in this Integrated Project, which will run from in its first phase. The views and conclusions contained herein are those of the authors and should not be interpreted as necessarily representing the s Project. The authors would like to thank all members of the s WP2, who have all contributed to the development of the architecture presented in this paper. REFERENCES [1] WWI s, [2] Niebert, N., Flinck, H. Hancock, R. Karl, H. Prehofer, C. s Research for Communication s Beyond 3G - IST Mobile and Wireless Communications Summit 2004 [3] G.P. Koudouridis, R. Agüero, E. Alexandri, J. Choque, K. Dimou, H.R. Karimi, H. Lederer, J. Sachs, R. Sigle. Generic Link Layer Functionality for Multi-Radio s, IST Mobile and Wireless Communications Summit 2005 [4] F. Berggren, I. Karla, R. Litjens, P. Magnusson, F. Meago, R. Veronesi, H. Tang., Multi-Radio Resource Management for Communication s Beyond 3G, VTC Fall [5] R. Agüero et al RRM Challenges for Non-Conventional and Low Cost s in s, submitted to WPMC 2005 [6] 3GPP TR Improvement of Radio Resource Management (RRM) across RNS and RNS/BSS. [7] A. Furuskär, J. Zander, Multiservice allocation for multiaccess wireless systems, IEEE Trans. Wireless Comm, vol. 4, no. 1, [8] A. Tölli, P. Hakalin,, H. Holma, Performance of common radio resource management (CRRM), IEEE ICC [9] F. Malvasi et al., Traffic control algorithms for a multi access network scenario comprising GPRS and UMTS, VTC Spring [10] J.Sachs, "A Generic Link Layer for Future Generation Wireless ing,", IEEE ICC'03, May [11] M.Siebert, B.Walke, Design of Generic and Adaptive Protocol Software (DGAPS), Third Generation Wireless and Beyond 2001 (3Gwireless '01)